projects

Six builds, one per module.

Each project is a guided, step-by-step build on real data that uses everything its module taught. A project opens once its module's tests and practicals are solved.

projectmoduledatasteps

Fit a line by gradient descentPredict an iris petal's width from its length, training a model from scratch with the tools from this module.01 The maths of learningIris, 150 flowers6→A flower classifier from scratchClassify all three iris species two ways, k-nearest neighbours and softmax regression, and compare them honestly on held-out flowers.02 Supervised learningIris, 150 flowers, 3 species5→An honest diagnosis modelPredict whether a breast tumour is malignant from biopsy measurements, choosing the model with cross-validation and opening the test set exactly once.03 Models that generaliseBreast Cancer Wisconsin, 569 biopsies5→A digit recogniser from scratchBuild and train a neural network on handwritten digits with nothing but numpy: forward pass, backpropagation, gradient check, mini-batch training.04 Neural networksHandwritten digits, 1,797 images of 8×8 pixels5→Train a convolutional network from scratchTurn convolution into a matrix multiply, then train a small CNN on handwritten digits and look at the filters it learns.05 Deep architecturesHandwritten digits, 8×8 images5→A tiny language modelBuild a character-level language model of first names: count it, train it as a neural network, sample new names, and give it more context.06 Language models and generative AI778 common first names6→